Immune cell screening of a nanoparticle libraryimproves atherosclerosis therapyJun Tanga,b, Samantha Baxtera, Arjun Menonc, Amr Alaarga,d, Brenda L. Sanchez-Gaytana, Francois Faya, Yiming Zhaoa,Mireille Ouimetc, Mounia S. Brazaa, Valerie A. Longoe, Dalya Abdel-Attib, Raphael Duivenvoordena, Claudia Calcagnoa,Gert Stormd,f, Sotirios Tsimikasg, Kathryn J. Moorec, Filip K. Swirskih, Matthias Nahrendorfh, Edward A. Fisherc,Carlos Pérez-Medinaa, Zahi A. Fayada, Thomas Reinerb, and Willem J. M. Muldera,i,1

aTranslational and Molecular Imaging Institute, Icahn School of Medicine at Mount Sinai, 1470 Madison Avenue, New York, NY 10029; bDepartment ofRadiology, Memorial Sloan Kettering Cancer Center, 1275 York Avenue, New York, NY 10065; cDivision of Cardiology, Department of Medicine, Marc andRuti Bell Program in Vascular Biology, New York University School of Medicine, 550 First Avenue, New York, NY 10016; dDepartment of Biomaterials Scienceand Technology, MIRA Institute for Biomedical Technology and Technical Medicine, University of Twente, 7500 AE Enschede, The Netherlands; eSmall-Animal Imaging Core Facility, Memorial Sloan Kettering Cancer Center, 1275 York Avenue, New York, NY 10065; fDepartment of Pharmaceutics, UtrechtInstitute of Pharmaceutical Sciences, Utrecht University, 3584 CG Utrecht, The Netherlands; gSulpizio Cardiovascular Center, Department of Medicine,University of California, San Diego, 9434 Medical Center Drive, La Jolla, CA 92037; hCenter for Systems Biology, Massachusetts General Hospital and HarvardMedical School, Simches Research Building, 185 Cambridge Street, Boston, MA 02114; and iDepartment of Medical Biochemistry, Academic Medical Center,Meibergdreef 9, 1105 AZ Amsterdam, The Netherlands

Edited by Liangfang Zhang, University of California, San Diego, La Jolla, CA, and accepted by Editorial Board Member Mark E. Davis September 12, 2016(received for review June 14, 2016)

Immunological complexity in atherosclerosis warrants targetedtreatment of specific inflammatory cells that aggravate the disease.With the initiation of large phase III trials investigating immuno-modulatory drugs for atherosclerosis, cardiovascular disease treat-ment enters a new era. We here propose a radically differentapproach: implementing and evaluating in vivo a combinatoriallibrary of nanoparticles with distinct physiochemical properties anddifferential immune cell specificities. The library’s nanoparticles arebased on endogenous high-density lipoprotein, which can preferen-tially deliver therapeutic compounds to pathological macrophages inatherosclerosis. Using the apolipoprotein E-deficient (Apoe−/−)mouse model of atherosclerosis, we quantitatively evaluated thelibrary’s immune cell specificity by combining immunological tech-niques and in vivo positron emission tomography imaging. Based onthis screen, we formulated a liver X receptor agonist (GW3965) andabolished its liver toxicity while still preserving its therapeutic func-tion. Screening the immune cell specificity of nanoparticles can beused to develop tailored therapies for atherosclerosis and otherinflammatory diseases.

nanomedicine | drug delivery | immunotherapy | molecular imaging |atherosclerosis

Research in the past decades has revealed the immune system’scentral role in the pathophysiology of cancer (1), diabetes (2),

and atherosclerosis (3, 4). Because macrophages drive pathologicalprogression of these diseases, immunomodulatory small-moleculecompounds modulating macrophage function are promising ther-apeutic candidates for treating these maladies. However, thesecompounds’ lack of cellular specificity necessitates a strategy fortargeted delivery to harmful immune cells without negatively af-fecting beneficial immune cells. Despite numerous studies ofnanoparticle-based delivery, rational attempts to screen me-ticulously designed nanoparticles for immune cell specificityin vivo have never been reported.We created a combinatorial library of hybrid lipoprotein-

inspired nanoparticles with distinct physiochemical properties(particularly size and chemical composition) with differentialimmune cell specificity. We then chose atherosclerosis, a lipid-driven inflammatory process of the large arteries, as a modeldisease to evaluate our nanoparticle library. Atherosclerosisaccounts for the majority of cardiovascular deaths worldwide(5), and macrophages are the major immune cells to drive thepathological inflammation and the progression of atheroscle-rotic plaques (6, 7). These plaques, which are present throughoutthe vasculature, have a complicated cellular composition (8). To

improve the therapeutic index of small-molecule immunomodu-latory compounds, plaque macrophage-specific delivery, withminimal delivery to nonpathological cells in healthy tissues, isessential.We used the apolipoprotein E-deficient (Apoe−/−) mouse model

of atherosclerosis to evaluate our nanoparticle library, because itaccurately recapitulates some important immunological aspects ofhuman atherosclerosis (9). Using a combination of optical methods,immunological techniques, and in vivo positron emission tomog-raphy (PET) imaging, we carefully selected candidate nanoparticlesfrom the library for subsequent atherosclerosis drug delivery studies.As a proof of concept, we incorporated the liver X receptor agonistGW3965, a therapeutic compound that did not reach clinical ap-plication due to its serious adverse effects on the liver (10, 11), in twonanoparticles—one with favorable organ distribution and immunecell specificity and one without. In Apoe−/− mice with advanceddisease, the favorable nanoparticle from the library screen was

Significance

The immune system plays an essential role in the pathophysiol-ogy of major diseases such as atherosclerosis, diabetes, andcancer, which has inspired the development of numerous smallmolecules to modulate immune cells, intending to create immu-notherapies for these diseases. Tissue- and cell-specific delivery ofthese small molecules is the key to transform these compoundsto safe, potent immunotherapies. In this study, we present an invivo nanoparticle screen approach that involves designing andevaluating a library of nanoparticles with distinct immune celltargeting specificity. This study carries out a systematic in vivoimmune cell screening to create effective nanoparticle-basedimmunotherapy for modulating the pathological immune cellsin atherosclerosis.

Author contributions: J.T., G.S., S.T., K.J.M., F.K.S., M.N., E.A.F., C.P.-M., Z.A.F., T.R., andW.J.M.M. designed research; J.T., S.B., A.M., A.A., B.L.S.-G., F.F., Y.Z., M.O., M.S.B., V.A.L.,D.A.-A., and C.P.-M. performed research; J.T., C.C., S.T., K.J.M., E.A.F., Z.A.F., T.R., andW.J.M.M. contributed new reagents/analytic tools; J.T., A.M., B.L.S.-G., M.O., R.D., F.K.S.,M.N., E.A.F., C.P.-M., T.R., and W.J.M.M. analyzed data; J.T., S.B., C.P.-M., and W.J.M.M.wrote the paper; and W.J.M.M. supervised the whole project.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission. L.Z. is a Guest Editor invited by the EditorialBoard.1To whom correspondence should be addressed. Email: Willem.mulder@mssm.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1609629113/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1609629113 PNAS | Published online October 17, 2016 | E6731–E6740

ENGINEE

RING

IMMUNOLO

GYAND

INFLAMMATION

PNASPL

US

Dow

nloa

ded

by g

uest

on

Dec

embe

r 31

, 202

0

shown to abolish GW3965’s liver toxicity while remaining ef-fective on atherosclerotic plaque macrophages.

ResultsStudy Design. We created a combinatorial library comprising 15high-density lipoprotein-mimicking nanoparticles and two exten-sively studied nanoparticles, a PEGylated micellar and a long-circulating liposomal nanoparticle. We then combined in vitroassays and in vivo experiments in atherosclerotic Apoe−/− mice tostudy the nanoparticle library’s biological behavior by using near-infrared fluorescence imaging (NIRF), flow cytometry, immuno-fluorescence, and radiolabeling. Based on the results of this libraryscreening, we formulated two GW3965-loaded HDL nanoparticles

with distinctly different immune cell specificity and organ distri-bution. Finally, we quantitatively studied the pharmacokinetics,immune cell specificity, liver toxicity, and therapeutic effects ofthese drug-loaded nanoparticles in Apoe−/− mice with advancedatherosclerosis (Fig. 1A).

HDL Nanoparticle Library. Previous studies indicate that the size,phospholipid composition, ratio of phospholipid to apolipo-protein A-1 (APOA1), or the inclusion of payloads can affectHDL-mimicking nanoparticles’ in vivo performance (12–14). Inour current study, we created a library containing HDL-mim-icking nanoparticles that differ in size, shape, composition, andpayload, all of which have been reported to affect nanoparticles’

A

C

Dia

met

er (

nm) 400

20080

40

0NP1 NP2 NP3 NP4 NP5 NP6 NP7 NP8 NP9 NP10 NP11 NP12 NP13 NP14 NP15 NP16 NP17

50

0

100

150

NP1 NP2 NP3 NP4 NP5 NP6 NP7 NP8 NP9 NP10 NP11 NP12 NP13 NP14 NP15 NP16 NP17

D

B

NP1 NP2 NP3 NP4 NP5 NP6 NP7 NP8 NP9 NP10 NP11 NP12 NP13 NP14 NP15 NP16 NP17

+GW3965

Nanoparticle in vivo immune cell screen Improving drug therapeutic index

Library creation Immunological screening Rational nanoparticle design In vivo PET-CT evaluation evaluation

Apoe-/-

!!

!!!

Fig. 1. Study design and in vitro characterization of the nanoparticle library. (A) In the in vivo immune cell screen study (left), we first created a combinatorialnanoparticle library and then evaluated the library in atherosclerotic Apoe−/− mice by using blood half-life determination, NIRF, and flow cytometry. The libraryscreening data lead to rational design of one GW3965-loaded nanoparticle with favorable characteristics and one without. In the second study, the twonanoparticles were radio- and fluorescence-labeled, and they were quantitatively and therapeutically evaluated by using PET-CT imaging, mRNA profiling, flowcytometry, and liver toxicity assays (right). (B) Representative high-magnification TEM images of negatively stained nanoparticles. Low-magnification images anda discussion of the different structures are presented in Fig. S1. (Scale bar, 50 nm.) (C) The size of nanoparticles as measured by dynamic laser scattering (DLS).(D) Cholesterol efflux capacity of the nanoparticles in primary macrophages normalized to natural human HDL (n = 6). Error bars are SDs. Color-coded bar at thebottom shows the relative rank of each nanoparticle, with the red indicating high cholesterol efflux efficiency and the blue indicating low efficiency.

E6732 | www.pnas.org/cgi/doi/10.1073/pnas.1609629113 Tang et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 31

, 202

0

in vivo targeting efficiency (15). To fine-tune nanoparticle sizeand morphology, we added either triglyceride or polymers [poly-lactic-co-glycolic acid (PLGA) or polylactic acid (PLA)] to theHDL core (Table S1); this allowed us to modulate nanoparticlesize from about 10 nm (NP1, NP2, NP3, and NP9) or 30 nm(NP6, NP7, NP10, and NP11) to over 100 nm (NP12) (Fig. 1 B

and C). We observed that the inclusion of a core component,namely triglycerides or polymers, results in a nanoparticle shapechange from discoidal to spherical, as can be clearly appreciatedwhen comparing NP5 to NP7 and NP15. In addition to size andshape, we also varied phospholipid composition (Table S1).Because oxidization greatly affects HDL function (16), we

B

A NP1NP2NP3

NP4NP5

NP6

NP7

NP8NP9

NP10

NP11

NP12

NP13

NP14

NP15

100

50

0

Rel

ativ

e pl

asm

a co

ncen

trat

ion

(%)

0 4 8 12 16 20 24

RankLo Hi

NP16

NP17

0 4 8 12 16 20 240 4 8 12 16 20 24

8

NP1 NP2 NP3 NP4 NP5 NP6 NP7 NP8 NP9 NP10 NP11 NP12 NP13 NP14 NP15 NP16 NP17

4.0

3.0

2.0

1.0

Aorta

Liver

Spleen

RankLo Hi

NP1 NP2 NP3 NP4 NP5 NP6 NP7 NP8 NP9 NP10 NP11 NP12 NP13 NP14 NP15 NP16 NP171.4

1.0

0.6

0.20

1

Total signal

Aorta-to-liver

Ao-to-sp

Ctrl

89 Zr-NP17

89 Zr-NP15

89 Zr-NP14

89 Zr-NP10

Acc

umul

atio

n in

ao

rtic

roo

t and

arc

h (%

ID)

0.10

0.06

0.04

0.02

0.00

E

0.08

**N.S.

*

C1.0

0.5

0.0Nor

mal

ized

blo

od r

adio

activ

ity

0 4 8 12 16 20 24 (hr)

89Zr-NP17

89Zr-NP15

89Zr-NP14

89Zr-NP10Aorta-to-liverAorta-to-spleen

Rel

ativ

e ac

cum

ulat

ion

(A.U

.) 0.3

0.2

0.1

0.0

89 Zr-NP17

89 Zr-NP15

89 Zr-NP14

89 Zr-NP10

89 Zr-NP17

89 Zr-NP15

89 Zr-NP14

89 Zr-NP10

Aorta

Liver

Spleen

D F

Fig. 2. In vivo evaluation of the nanoparticle library. (A) Relative nanoparticle plasma concentration in Apoe−/−mice that were fed a 12-wk high-cholesterol diet.The values were derived from the near-infrared dye DiR incorporated in the nanoparticles. Heat map below the graphs ranks the blood half-lives, with the redindicating a long and the blue a short blood half-life (n = 5 per nanoparticle). The values of blood half-lives are provided in Table S2. (B) Representative near-infrared fluorescence images of nanoparticle accumulation in aorta, liver, and spleen. The heat map below the aorta images ranks the mean total fluorescenttissue signal, the heat map below the liver images ranks the total aorta-to-liver signal, and the heat map below the spleen images ranks the mean aorta-to-spleenaccumulation (Ao-to-sp) ratio, with the red indicating a high and the blue a low ratio (n = 5 for each nanoparticle). Bar graphs are provided in Fig. S2. (C) Bloodhalf-lives of four selected nanoparticles radiolabeled with 89Zr (n = 3 per nanoparticle) were 7.0 h for 89Zr-NP10, 5.7 h for 89Zr-NP14, 7.1 h for 89Zr-NP15, and 6.6 hfor 89Zr-NP17. Blood radioactivity (percentage injected dose per gram of tissue) of all time points is normalized to that of the first time point—2 min after in-jection. (D) Representative autoradiography images of aortas, livers, and spleens 24 h after nanoparticle injection. Dashed windows indicate the aortic root andarch area analyzed in E. (E) Nanoparticle accumulation in aortic roots and arches as measured by percentage of injected dose (%ID) (n = 3 per nanoparticle).(F) Relative accumulation of nanoparticles between aortas and liver or spleen. Arbitrary units (A.U.) were defined by the aortic accumulation (percentage injecteddose per gram of tissue) divided by the hepatic or splenic accumulation (percentage injected dose per gram of tissue). Error bars are SDs. Statistics was calculatedwith nonparametric two-tailed Student’s t test. N.S., not statistically significant; *P < 0.05; **P < 0.01.

Tang et al. PNAS | Published online October 17, 2016 | E6733

ENGINEE

RING

IMMUNOLO

GYAND

INFLAMMATION

PNASPL

US

Dow

nloa

ded

by g

uest

on

Dec

embe

r 31

, 202

0

sought to test if this modification also changed the HDL-mim-icking nanoparticles’ drug delivery capability. We oxidized thephospholipids and APOA1 of NP1 to produce NP2 (Fig. S1B).The nanoparticle sizes were measured by dynamic light scat-tering, and their morphologies were visualized by transmissionelectron microscopy (TEM) (Fig. 1 B and C). Micellar (17) andliposomal (18) nanoparticles are established lipid-based plat-forms and served as references in this study.

HDL Nanoparticle Library’s Cholesterol Efflux Efficiency. NaturalHDL facilitates reverse cholesterol transport, the intrinsic mech-anism that removes cholesterol from macrophages in atheroscle-rotic plaques and protects against the atherosclerosis (19, 20).Cholesterol efflux capacity indicates the nanoparticles’ biologicalsimilarity to native HDL. Therefore, we measured the 17 nano-particles’ ability to induce cholesterol efflux from cholesterol-laden primary macrophages. We found 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC)-based nanoparticles (NP8,NP9, NP10, and NP11) to be the most efficient in extractingcholesterol from macrophages; in contrast, polymer-core HDLnanoparticles produced the least cholesterol efflux (NP12, NP13,NP14, and NP15). Liposomal and micellar nanoparticles, withoutAPOA1 on their surface to bind the cholesterol efflux receptorsAbca1 and Abcg1, performed similarly to polymer-core HDLnanoparticles (Fig. 1D). These results demonstrate the essentialrole of APOA1, as well as the impact of phospholipid and corecomposition, in promoting cholesterol efflux. It has been proposedthat APOA1 changes conformation when interacting with itsspecific receptors to extract cholesterol (19). The artificial poly-meric cores of NP12, NP13, NP14, and NP15 might limit con-formational flexibility of APOA1, leading to impaired cholesterolefflux. On the other hand, POPC-based HDL nanoparticles dis-play less rigidity, are more similar to natural HDL, and thereforeresult in more efficient cholesterol efflux.

Nanoparticles’ Physiochemical Properties Affect Their in VivoBehavior. We i.v. injected the library’s 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindotricarbocyanine iodide (DiR)-labeled nanoparticlesinto atherosclerotic Apoe−/− mice. We found that NP1 and NP10,with diameters between 7 nm and 30 nm, exhibited the longest bloodhalf-lives of 5.0 h and 6.3 h, respectively. NP12 and NP14, with di-ameters larger than 70 nm, had the shortest blood half-lives of 0.71 hand 0.67 h, respectively (Fig. 2A). The difference between the lon-gest and the shortest blood half-life was almost 10-fold (Table S2).Using NIRF, we investigated nanoparticle accumulation in the

heart, aorta, lung, liver, spleen, kidney, brain, and muscle 24 h afteri.v. administration (Fig. 2B and Fig. S2A). Among all nano-particles, liver accumulation was generally the highest, followed byspleen, kidney, and lung accumulation (Fig. S2B). Because theaorta is the primary target tissue whereas the liver and spleen areclearance organs, we determined accumulation in the aorta rela-tive to these two organs (21). Our measurements showed a 3.4-folddifference between the highest and lowest aorta-to-liver accumu-lation ratios (NP5 vs. NP12, P < 0.01, Fig. S2D), and a 4.7-folddifference between the aorta-to-spleen accumulation ratios (NP5vs. NP12, P < 0.01, Fig. S2E). For the nanoparticles’ relativeperformance, see Fig. 2B. Distribution of DiR-labeled nano-particle in certain tissues is difficult to quantify in vivo due to thelimited penetrating depth of light and varying absorbance ratesamong tissues (22). Radiolabeled nanoparticles’ biodistributionand pharmacokinetics can be quantitatively measured. From ourinitial optical imaging screen, we selected NP10, NP14, NP15, andNP17 to be labeled with Zr through the hydrophobic chelatordeferoxamine (DFO)-C34, which serves as a surrogate for hydro-phobic payloads. In line with the optical imaging results, we observedthat the smaller nanoparticles (NP10 and NP15) had longer bloodhalf-lives than the larger ones (NP14 and NP17, Fig. 2C). We alsoobserved that the HDL-based nanoparticles (NP10, NP14, and

NP15), particularly NP10, accumulated more efficiently in athero-sclerotic plaques than the liposomal nanoparticle (Fig. 2 D and E).Overall, NP10 displayed the most favorable performance among thefour selected nanoparticles (Fig. 2 C−F).

Distinct Immune Cell Targeting Patterns Within the NanoparticleLibrary. Macrophages and monocytes are the key immune cellsthat drive atherosclerosis progression (4). In Apoe−/− atheroscle-rotic mice, these cells mainly reside in atherosclerotic plaques,spleen, blood, and bone marrow (4). Using a robust flow cytometryprocedure adapted from previous studies (23, 24), we were able toidentify macrophages, monocytes, and nonmyeloid immune cells(Lin+) in the aortas; macrophages, Ly-6Chi monocytes, dendriticcells (DCs), and neutrophils in the spleens; and Ly-6Chi monocytes,Ly-6Clo monocytes, DCs, and neutrophils in the blood (Fig. 3A).In the aortas, all HDL-mimicking nanoparticles efficiently tar-

geted macrophages and monocytes. The difference between thehighest and lowest nanoparticle accumulations was 5.7-fold inmacrophages (NP3 vs. NP17, P < 0.01) and 2.7-fold in monocytes(NP3 vs. NP7, P < 0.001; Fig. 3B). In the spleens, all nanoparticleshad the highest accumulation in macrophages, which do the bulk ofnanoparticle clearance, with a 3.8-fold difference between thehighest and lowest accumulation levels (NP12 vs. NP17, P < 0.01,Fig. 3B). In the blood, DCs and Ly-6Chi monocytes displayed thehighest nanoparticle association, with a 3.8-fold difference betweenthe highest and lowest association levels (NP10 vs. NP5, P < 0.01) inDCs and a 3.79-fold difference in Ly-6Chi monocytes (NP3 vs.NP14, P < 0.0001, Fig. 3B). In addition, the nanoparticles were farless effective in targeting Ly-6Clo monocytes, the patrolling mono-cytes in the blood, compared with Ly-6Chi monocytes (Fig. 3B).Although aortic macrophages are the main target of immuno-

modulatory nanoparticles, splenic macrophages, which clearnanoparticles from the blood and reduce their bioavailability toaortic macrophages (25), need to be avoided. We evaluated theratios of nanoparticle accumulation in aortic macrophages versussplenic macrophages, and we found a 3.8-fold difference betweenthe highest and lowest aortic-to-splenic ratios (NP16 vs. NP12, P <0.0001, Fig. S3A). This differential targeting specificity to athero-sclerotic macrophages was confirmed by immunofluorescence inthe aortic roots (Fig. S3B). Altogether, the flow cytometry datareveal that the distinct physiochemical nanoparticle propertieswithin the library lead to drastically different immune cell targetingpatterns (Fig. 3C).

GW3965-Loaded Nanoparticle Development. Liver X receptor(LXR) agonists promote cholesterol efflux from macrophages inatherosclerotic plaques, and they have been proposed as novelimmunomodulatory drugs for the disease (26). However, mostexperimental LXR agonists fail clinical translation or early-stageclinical trials due to poor safety profiles. For example, GW3965, aneffective LXR agonist promoting cholesterol efflux from athero-sclerotic macrophages (27, 28), did not reach the clinical phase,because of its liver toxicity in hamsters and monkeys (10), as well asin human hepatocytes (11).In our nanoparticle library studies, NP10 was found to have high

cholesterol efflux promotion efficiency, a long blood half-life, highrelative aorta-to-liver accumulation, and a high relative aortic-to-splenic macrophage association ratio (Fig. S4A). These featuresmake NP10 a promising candidate for avoiding GW3965’s livertoxicity and enhancing its efficacy on atherosclerotic plaque mac-rophages. Therefore, we replaced NP10’s hydrophobic triglyceridecargo with hydrophobic GW3965 and created a GW3965-loadednanoparticle (Rx-HDL) that was morphologically similar but notidentical, due to the different nanoparticle composition, to NP10in size (∼30 nm), phospholipid composition (POPC-dominant),and morphology (Fig. S4B). Further, we identified NP14 as anunfavorable nanoparticle for GW3965 delivery due to the nano-particle’s poor cholesterol efflux efficiency, short blood half-life,

E6734 | www.pnas.org/cgi/doi/10.1073/pnas.1609629113 Tang et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 31

, 202

0

and low relative aorta-to-liver accumulation (Fig. S4A). Byloading GW3965 into the PLGA matrix of NP14, we created aPLGA-core GW3965-loaded nanoparticle (Rx-PLGA-HDL)with similar size, phospholipid composition, and morphology toNP14 (Fig. S4B). Notably, the size and cholesterol efflux ca-pability of the two drug-loaded nanoparticles were drasticallydifferent (Fig. S4 B and C).

Quantitative Evaluation of GW3965-Loaded Nanoparticles. Nano-particles that are 89Zr-labeled can be quantitatively characterizedby in vivo PET imaging as well as by ex vivo radioactivitycounting (29). To accurately understand the in vivo perfor-mance of the two nanoparticles, we loaded the hydrophobic89Zr-DFO-C34 into Rx-HDL and Rx-PLGA-HDL (Fig. 4A),and radioactive high-performance chromatography showed

CD11b

Lin

Ly-6C

CD

11c

Ly-6C

SS

C

Cou

nt

NP

Blood

Neutrophil Ly-6Chi Mo Ly-6Clo Mo DC NeutrophilDC

Ly-6Chi MoLy-6Clo Mo

CD11b

Lin

CD

11c

F4/80 Ly-6C

SS

C

Spleen

Neutrophil DCM

Cou

nt

NP

Ly-6Chi Mo

M Ly-6Chi Mo DC Neutrophil

Lin+

CD11b Ly-6C

Lin

F4/

80

Aorta

M

Cou

nt

NP

M Lin+

Ly-6Chi Mo

Ly-6Chi Mo

A

BM

M

Ly-6Chi Mo Ly-6Clo Mo

DC

DC

Lin+

Neutrophil

Neutrophil

2

1

0

4

2

0

3

1.5

0

MF

I ( x

10,

000)

Aorta

Spleen

Blood

Ly-6Chi Mo

Ly-6Chi Mo

NP1 NP2 NP3 NP4 NP5 NP6 NP7 NP8 NP9 NP10 NP11 NP12 NP13 NP14 NP15 NP16 NP17

RankLo Hi

Aortic M

Blood Ly-6Chi MoSplenic M

Ao-to-sp M

C

Fig. 3. Nanoparticle immune cell specificity. (A) The flow cytometry gating procedures to identify relevant immune cells in aorta, spleen, and blood. Blackhistograms on the right show representative signal distribution of different immune cells in the mice injected with nanoparticles compared with the cells fromcontrol animals injected with PBS (gray histogram on the left in each graph). (B) Quantification of mean fluorescence intensity (MFI) of each immune cell typein different tissues; n = 5 for each nanoparticle, and error bars are SEMs. (C) Heat map ranks targeting efficiency in key immune cells (aortic macrophages,spleen macrophages, and blood Ly-6Chi monocytes), with red indicating a high and blue indicating a low MFI in the first three rows. The last row shows theaortic-to-splenic macrophage MFI ratio (Ao-to-sp MΦ), and its quantitative values are provided in Fig. S3A.

Tang et al. PNAS | Published online October 17, 2016 | E6735

ENGINEE

RING

IMMUNOLO

GYAND

INFLAMMATION

PNASPL

US

Dow

nloa

ded

by g

uest

on

Dec

embe

r 31

, 202

0

both Rx-HDL and Rx-PLGA-HDL to be efficiently radiola-beled (Fig. S4D).In Apoe−/− atherosclerotic mice, Rx-HDL circulated in the blood

much longer (weighted t1/2 = 10.5 h, n = 3) than Rx-PLGA-HDL(weighted t1/2 = 5.0 h, n = 3, Fig. 4B) or its precursor NP10 (t1/2 =7.0 h, Fig. 2C), demonstrating its favorable features. We then usedPET-computed tomography hybrid imaging (PET-CT) to measurethe nanoparticle dynamic accumulation in the cardiac blood pool,the liver, and the spleen 30 min and 24 h after i.v. administration(n = 5 per nanoparticle, Fig. 4C; also see Movies S1–S4). After 30min, Rx-HDL had higher accumulation in the cardiac blood pool(29.9 maximum percentage injected dose per gram of tissue (Max%ID/g) vs. 24.5 Max %ID/g, P < 0.05) but lower accumulationin the liver than Rx-PLGA-HDL (21.7 Max %ID/g vs. 29.7 Max%ID/g, P < 0.05). After 24 h, Rx-HDL liver accumulation was still36% lower than Rx-PLGA-HDL (26.5 Max %ID/g vs. 41.7 Max%ID/g, P < 0.05, Fig. 4 C and D). Autoradiography revealed thatboth nanoparticles displayed patchy aorta accumulation, in accor-dance with the heterogeneous distribution of atherosclerotic pla-ques in this tissue (30) (Fig. 4E). Additionally, autoradiography

confirmed the highest nanoparticle accumulation to be in the liverand spleen (Fig. 4E), a result that was corroborated by an extensivebiodistribution analysis (Fig. S4E).Having labeled both nanoparticles with DiR (Fig. 4A), we used

our flow cytometry protocol (Fig. 3A) to quantify immune cell tar-geting specificity (Fig. S4F). Rx-HDL predominantly targeted mac-rophages in the aortas, and its accumulation was twofold higher thanRx-PLGA-HDL (P < 0.01). In the spleen, Rx-HDL had 33% lessaccumulation in splenic macrophages than Rx-PLGA-HDL (P <0.05). In the blood, Rx-HDL targeted Ly-6Chi monocytes 2.4-foldmore efficiently than Rx-PLGA-HDL (P < 0.01, Fig. 4 F and G).Collectively, these data show that, compared with Rx-PLGA-HDL,Rx-HDL has a longer blood half-life, lower accumulation in theliver, higher accumulation in atherosclerotic plaque macrophages,and lower accumulation in splenic macrophages.

Nanoparticle Abolishes Liver Toxicity and Preserves Efficacy ofGW3965. To test if Rx-HDL’s optimal in vivo performance re-duced GW3965s liver toxicity, we gave four i.v. injections (oneinjection every 2 d, at a dose of 10 mg/kg GW3865) of Rx-HDL, its

60

40

0

20Max

%ID

/g

24 hRx-HDLRx-PLGA-HDL

Heart Liver Spleen

0.5 h

* *

*

E

Rx-HDL Rx-PLGA-HDL

24 h0.5 h

Rx-HDL Rx-PLGA-HDL

HeLiSp

GW3965 ApoA1

POPC/PHPC

DMPC/MHPC

PLGA

Rx-HDL Rx-PLGA-HDL

89Zr-C34 / DiR

Rx-

HD

L

Aorta Lung Liver Spleen Kidney Muscle Heart

Rx-

PLG

A-H

DL

4

M

Ly6C

hi Mo

Lin+

Ly6C

hi Mo

DC

Neutro

phil

M

Ly6C

hi Mo

Ly6C

lo Mo

DC

Neutro

phil

Aorta Spleen Blood

3

2

1

0

MF

I (x

104 ) Rx-HDL

Rx-PLGA-HDL

***

**

*

**

**

40

20

00 4 8 12 16 20 24

%ID

/g

Rx-HDL

Rx-PLGA-HDL

hour

G

Cou

nt

NP

Rx-PLGA-HDLRx-HDL

AorticM

Splenic M Blood Ly6Chi Mo

D E

CBA

F

60

Fig. 4. In vivo quantitative evaluation of GW3965-loaded nanoparticles. (A) Schematic depictions of small Rx-HDL plaque macrophage-targeting and largepolymer-hybrid Rx-PLGA-HDL nanoparticles. The two nanoparticles were either radiolabeled with 89Zr or labeled with the near-infrared fluorescent dye DiR.(B) Blood half-lives were determined in three mice per nanoparticle. (C) Representative PET images of mice that received either the small or large nanoparticleat 0.5 h and 24 h after i.v. administration. The 3D-rendered images are provided as Movies S1–S4 (n = 5 per nanoparticle). (D) Quantification of radioactivity inthe heart, liver, and spleen. (E) Representative autoradiographic images of key organs. Full biodistribution of all organs is provided in Fig. S4E. (F) Repre-sentative histograms of selected immune cell targeting specificity in the aorta, spleen, and blood. (G) MFI quantification of relevant immune cells in the tissues(n = 4 per nanoparticle). Error bars are SEM. Statistics was calculated with nonparametric two-tailed Student’s t test. *P < 0.05; **P < 0.01.

E6736 | www.pnas.org/cgi/doi/10.1073/pnas.1609629113 Tang et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 31

, 202

0

vehicle control (HDL), Rx-PLGA-HDL, its vehicle control(PLGA-HDL), and PBS to Apoe−/− atherosclerotic mice (n = 12per group).In the liver, Rx-PLGA-HDL increased the expression of two of

the three major GW3965-related toxicity genes, and Rx-HDL in-creased the expression of one gene (Fig. 5A). Importantly, wemeasured hepatic triglyceride and cholesterol levels, which are themajor biomarkers of GW3965-induced hepatic steatosis (10). TheRx-PLGA-HDL group had 35.4% more hepatic triglyceride than itscontrol vehicle PLGA-HDL group (P = 0.014) and 21% more thanthe PBS group (P = 0.12). For hepatic cholesterol, the Rx-PLGA-HDL group had 31% higher levels than the PLGA-HDL group (P =0.018) and 17% higher levels than the PBS group (P = 0.014). Theseresults suggest that the high liver accumulation of Rx-PLGA-HDLcaused severe liver toxicity (Fig. 5 B and C). On the other hand, theRx-HDL group had 26.5% lower hepatic triglyceride levels than itsvehicle HDL control group (P = 0.06) and 22.7% lower levelsthan the PBS group (P = 0.076). The Rx-HDL group also had

20% lower hepatic cholesterol levels than the HDL group (P =0.06) and 33.3% lower levels than the PBS group (P = 0.00043).A recent study suggested that high HDL levels are associatedwith a lower degree of steatosis, which might explain the re-duced hepatic triglyceride and cholesterol levels in mice treatedwith vehicle HDL nanoparticles (31). Furthermore, GW3965has been reported to increase HDL levels in mice (32), likelyexplaining the additional hepatic benefits in Rx-HDL-treatedmice. Most importantly, compared with the Rx-PLGA-HDLgroup, the Rx-HDL group had 36.1% lower hepatic triglyceride(P = 0.00083) and 43.0% lower hepatic cholesterol levels (P <0.0001). These results demonstrate that the two GW3965-loaded nanoparticles’ distinct liver accumulations resulted indifferential liver toxicity profiles (Fig. 5 A−C).To measure the treatments’ efficacies on aortic macrophages, we

isolated the cells from aortic roots by laser capture microdissectionand measured their mRNA expression levels by quantitative real-time PCR (qPCR). We found that Rx-HDL increased expression

D

Ccl2

4211/21/4

Ccr2Ica

m1

Ifng

Il12a

Il1a

Il1b

Il6 Nos2M

mp2

Mm

p9

TnfaVca

m1

Cd206

Foxp3

Il10

Mrc

2Tgif

1Ccr

7

E

Abca1

Abcg1

Nr1h2

Nr1h3

Srebp

1c

GW3965 target genes

Rx-PLGA-HDL

Rx-HDL

S

3211/21/3

A B 6

4

2

0

Hep

atic

trig

lyce

ride

(g/

mg

prot

ein)

P < 0.001

PBSHDL

PLGA-H

DL

Rx-HDL

Rx-PLG

A-HDL

Hep

atic

cho

lest

erol

(g/

mg

prot

ein)

40

30

20

10

0

P < 0.0001

PBSHDL

PLGA-H

DL

Rx-HDL

Rx-PLG

A-HDL

Abca1

Abcg1

Srebp

1c

1.21.110.90.8

Rx-PLGA-HDLRx-HDL

A

Liver toxicity genes C

F

Rx-HDLPBS

Cel

lula

r lip

id le

vels

(x 1

08 A

U)

12

8

4

0

P = 0.225

Rx-HDLPBS

Cel

lula

r lip

id le

vels

(x

108 A

U)

1.5

1.0

0.5

0.0

P = 0.011G

Rx-HDLPBS

Cel

lula

r lip

id le

vels

(x 1

08 A

U)

90

60

30

0

P = 0.021HAortic macrophage Aortic monocyte Aortic non-immune cell

Fig. 5. Evaluation of the toxicity and efficacy of GW3965 (Rx)-loaded nanoparticles. In A−E, Apoe−/− mice received four i.v. administrations of nanoparticlesor PBS on every other day (n = 12 per group). (A) The mRNA expression levels of three GW3965 target genes in liver homogenates. (B) Triglyceride and(C) cholesterol levels in the liver, the primary organ suffering from the toxic effects of GW3965. (D) The mRNA expression levels of five GW3965 target genesin aortic macrophages. (E) The mRNA expression levels of 19 inflammation-related genes in aortic macrophages. All gene expression was normalized tohousekeeping gene Hprt1. The bar graph presentation of the heat maps is in Fig. S5 (n = 12 per group). In F−H, Apoe−/− mice (n = 10 in PBS; n = 9 in Rx-HDL)received 12 i.v. injections in 6 wk. Lipid levels of aortic cells were analyzed by flow cytometry. Cellular lipid levels were calculated by multiplying the meanfluorescence intensity of a cell type by the number of the cells per aorta (AU = MFI × number of cells). The total lipid levels of aortic (F) macrophages,(G) monocytes, or (H) nonimmune CD45− cells per aorta in the mice are presented here. Gating procedure is provided in Fig. S6 A and B. Error bars in all graphsare SEM. Statistics was calculated with nonparametric two-tailed Student’s t test.

Tang et al. PNAS | Published online October 17, 2016 | E6737

ENGINEE

RING

IMMUNOLO

GYAND

INFLAMMATION

PNASPL

US

Dow

nloa

ded

by g

uest

on

Dec

embe

r 31

, 202

0

of five GW3965 target genes compared with the vehicle HDLcontrol, whereas Rx-PLGA-HDL increased only four genes com-pared with the vehicle PLGA-HDL control (Fig. 5D). Further-more, Rx-HDL produced elevated expression of the five targetgenes compared with Rx-PLGA-HDL (Fig. S5 A and B). Theseresults suggest successful GW3965 delivery to macrophages inatherosclerotic plaques. We did not observe clear effects of eithernanoparticle on genes related to macrophage inflammation, as thenumber of genes with increased expression was almost equal to thenumber with decreased expression (Fig. 5E and Fig. S5 A and B).Similarly, the macrophage levels in aortic roots were the sameamong all groups (Fig. S5 D and E).Based on the favorable properties of Rx-HDL, we further

evaluated its therapeutic effects in a 6-wk, long-term treatmentregimen focusing on cellular lipid levels in aortas. Working from aBODIPY (4,4-difluoro-1,3,5,7,8-pentamethyl-4-bora-3a,4a-diaza-s-indacene)-based flow cytometry protocol (33), we developed aprocedure to quantify the cellular lipid levels of macrophages,monocytes, and CD45 negative nonimmune cells (CD45−) in theaortas (Fig. S6). Importantly, we found that the 6-wk treatment ofRx-HDL (10 mg/kg GW3965, two i.v. injections per week) resultedin 28% less total lipid in aortic macrophages (P = 0.224; Fig. 5F).Further, the treatment reduced total lipid levels in monocytes by43.0% (P = 0.011, Fig. 5G) and in all CD45− nonimmune cells by40.0% (P = 0.021, Fig. 5H). Furthermore, no therapeutic effectswere observed when the compound was given orally with the same6-wk treatment regimen (10 mg/kg GW3965, two gavages perweek) compared with the placebo-treated group (Fig. S6 C−H).Consistent with the liver toxicity results after the short-term Rx-HDL treatment, blood cholesterol and triglyceride levels weresimilar to those in mice receiving PBS treatment (Fig. S7 A and B).In addition, the long-term treatment of Rx-HDL did not cause anyobservable toxicity in the liver, kidney, heart, and blood cells(Fig. S7 C−L). Taken together, the results show that long-termRx-HDL treatment produces significant therapeutic benefitswithout causing toxicity in major organs, most notably the liver,whereas long-term oral treatment at the same dose did notproduce any therapeutic benefits.In summary, we developed a rational library screen strategy to

identify nanoparticles with favorable immune cell specificity andbiodistribution in an atherosclerosis mouse model. On the basis ofthis nanoparticle screen, we optimized GW3965 delivery to plaquemacrophages and, while preserving its efficacy on atheroscleroticplaques (Fig. 5 D−H), abolished GW3965 liver toxicity (Fig. 5 A−Cand Fig. S7), a well-known adverse effect of LXR agonists.

DiscussionBy fine-tuning the components and synthesis procedures, we createda combinatorial library of 17 nanoparticles with distinct composi-tion, size, and morphology. These distinct physiochemical propertiesresulted in an approximate sixfold difference in promoting choles-terol efflux from macrophages, 10-fold difference among blood half-lives, 3.4-fold difference in relative aorta-to-liver accumulation, and3.8-fold difference in relative aortic-to-splenic macrophage accu-mulation. From this library screening, we identified the favorablelipid composition (POPC-dominant), pharmacokinetics (long bloodhalf-life), size (around 30 nm), and morphology (spherical) toachieve optimal plaque macrophage-specific drug delivery. We hy-pothesize that the combination of long blood half-life and small sizeallow efficient and prolonged atherosclerotic plaque penetrationand subsequent macrophage accumulation. A favorable nanoparticlelipid composition and morphology increases stability and promotesdelivery of the encapsulated small molecules to the targeted cells, assuggested by our recent study (34). In a proof-of-concept applica-tion, we used these guidelines to identify two nanoparticles from thelibrary as the most and least favorable nanoparticles for deliveringthe liver-toxic compound GW3965. Although the unfavorablenanoparticle Rx-PLGA-HDL caused severe GW3965-induced liver

toxicity, the favorable nanoparticle Rx-HDL did not cause observ-able liver toxicity in treated animals but did preserve the compound’stherapeutic efficacy on the atherosclerotic plaques.Despite the clinical introduction of antibody immunotherapies

for atherosclerosis (35), such biological drugs can only modulate alimited number of extracellular targets, such as PCSK9 (36, 37) orreceptors on cell surface (38). Intracellular entities present manymore immunological targets that can be effectively controlled byimmunomodulatory small molecules. Most experimental smallmolecules for atherosclerosis, however, failed clinical trials due totheir unfavorable toxicity profiles, generally caused by high accu-mulation in nontargeted tissues or in nontargeted cells within thetargeted tissues (39). To convert these immunomodulatory smallmolecules into precision medicines for atherosclerosis, organ-specific and cell-specific delivery is highly desirable.Our approach allows the creation and immunological screening

of a combinatorial nanoparticle library with distinct organ bio-distribution and immune cell targeting specificities. By using PETimaging, NIRF, and flow cytometry to generate an extensive da-tabase detailing the in vivo performance of the library nano-particles, we were able to rationally design a strategy that avoidsthe specific limitations of an immunomodulatory compound. Throughthis process, we successfully converted a liver-toxic immunomod-ulatory compound into a precision nanomedicine for atheroscle-rotic plaque macrophage treatment.A polymer-core nanoparticle (NP13) in the library has similarly

optimal performance to the best-performing NP10 (Fig. S4A).Polymer-based nanoparticles have been formulated with chemicalcompounds (40), peptides (41), and nucleic acids (42). This varietysuggests that NP13 may be able to deliver a wide range of therapeuticmolecules. In addition, a few nanoparticles (NP3, NP9, and NP10) inthe library show high targeting specificity to Ly-6Chi monocytes andDCs, which are attractive targets in certain types of cancer (43),asthma (44), and diabetes (45). It would be interesting to apply thesame library screening strategy to develop nanoparticle-based specificdrug delivery to these immune cells in relevant diseases.This nanoparticle library screening strategy in immune cells

allowed us to improve the therapeutic index of an immunomod-ulatory molecule that causes hepatic toxicity and has failed clinicaltranslation. Our precision nanomedicine strategy is radically dif-ferent from the current clinical therapeutics as well as those inexperimental phases. Moreover, the approach’s potential to delivervarious compounds preferentially to other immune cells wouldexpand its application to numerous immunologically implicateddiseases, such as myocardial infarction, diabetes, and cancer.

Materials and MethodsSynthesis of Library Nanoparticles. The compositions of all NP synthesis ma-terials are listed in Table S1. The synthesis procedure for NP1 through NP11 wassimilar to a previous method (12). Briefly, phospholipids, DiR, and triglyceridewere dissolved in a chloroform/methanol solvent, dried to form thin film, andthen hydrated with human APOA1 solution. The homogenized solution wassonicated with a tip sonicator, and the aggregates and free lipids were re-moved by passing through a series of filters. The synthesis procedure for NP12through NP15 was adapted from a previous microfluidics-based method (14).Briefly, a solution containing 0.79 mL of a PLGA or PLA solution in acetonitrile(100 mg/mL), 1.58 mL of a 3:1 molar ratio of 1,2-dimyristoyl-sn-glycero-3-phosphocholine/1-myristoyl-2-hydroxy-sn-glycero- phosphocholine (DMPC/MHPC) in ethanol (5 mg/mL), 0.36 mg of DiR, 7.9 mL of ethanol, and 14.5 mL ofacetonitrile was prepared. The aforementioned solution was injected in themiddle channel of the microfluidic device at a rate of 2 mL/min, and a solutionof APOA1 (0.01 mg/mL in PBS) was injected in the outer channels at a rate of10 mL/min. The product was collected, washed with PBS, and concentratedusing tangential filtration [100,000-Da molecular mass cutoff (MMCO)] toremove acetonitrile, ethanol, and lipid-free APOA1. For micelle NP16, 2 mL ofchloroform solution containing 211 mg of 1,2-distearoyl-sn-glycero-3-phos-phoethanolamine-N-[amino(polyethylene glycol)-2000] (DSPE-PEG2000) and0.4 mg of DIR (0.5%mol) was slowly dripped into 10mL of PBS solution heatedat 85 °C under vigorous stirring. After dripping, the solutionwas kept at 85 °C untilall of the chloroformwas totally evaporated. Themicelle solution was washed and

E6738 | www.pnas.org/cgi/doi/10.1073/pnas.1609629113 Tang et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 31

, 202

0

concentrated in PBS using Millipore centrifugal filter (50,0000 Da MMCO). For li-posome NP17, a lipid film was first prepared by evaporating a chloroform solutioncontaining 35.4 mg of 1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC),11.1 mg of DSPE-PEG2000, 10.24 mg of cholesterol, and 0.4 mg of DiR (61.1%,2%, 33.4%, 0.5% in molar percentage). The residue of chloroform was re-moved by blowing nitrogen gas. The resulting film was hydrated with 10 mLPBS, vortexed, and subsequently sonicated for 25 min. After centrifugation at18 × g for 10 min to remove aggregates, the liposome solution was washedand concentrated in PBS using Millipore centrifugal filter (100,000 DaMMCO).To oxidize NP1, a PBS solution containing NP1 was stirred at 37 °C in thepresence of EDTA, α-tocopherol, and 2,2’-Azo-bis(2-amidinopropane) dihy-drochloride for 20 h. The resulting oxidized nanoparticle (NP2) was washedthoroughly with PBS.

Synthesis of GW3965-Loaded Nanoparticles. The synthesis procedure for Rx-HDLwas similar to that used for NP10. In addition to phospholipids [POPC and1-palmitoyl-2-hexadecyl-sn-glycero-3-phosphocholine (PHPC), 3:1 by weight] andDiR, GW3965 was added to make up 26% by weight of the total starting ma-terials. For Rx-PLGA-HDL, a solution containing 1 mL of GW3965 in DMSO(50 mg/mL), 2 mL of PLGA (100 mg/mL), 2 mL of lipids (10 mg/mL), 53 mL ofacetonitrile, and 22 mL of ethanol was prepared. The organic solution was in-jected in the middle channel of the microfluidic device at a rate of 2 mL/min,while a solution of APOA1 (0.01 mg/mL in 1× PBS) was injected in the outerchannels at a rate of 10mL/min. The product was collected, washedwith PBS, andconcentrated to ∼2 mg/mL using tangential and centrifugal filtration (100,000 DaMMCO and 300,000 Da MMCO, respectively). After synthesis, GW9365 wasextracted from the final nanoparticles by acetonitrile, and the compound amountwas measured by HPLC. The incorporation rate of Rx-HDL was above 90%, andthat for Rx-PLGA-HDL was about 50%. The GW3965 injecting dose was adjustedon the basis of themeasured GW3965 concentration in the nanoparticle solution.

Animals and Treatment Procedure. All procedures and experiments were ap-proved by the Institutional Animal Care and Use Committee of Icahn School ofMedicine at Mount Sinai. About 300 female Apoe−/− mice (B6.129P2Apoetm1Unc/J)were purchased from Jackson Laboratories and then fed a high-fat diet(Harlan Teklad TD.88137, 42% calories from fat) for 16 wk. These mice de-veloped advanced atherosclerosis in their aortic roots and aortas after 10 wkof high-fat diet and mimicked advanced disease in humans after 16 wk of thediet. Female C57BL/6 mice were fed on regular chow. All nanoparticles werei.v. injected into the lateral tail veins. Care was taken to ensure that each in-jection was less than 150 μL in volume. For the in vivo characterization study ofthe 17 nanoparticles, Rx-HDL, and Rx-PLGA-HDL, one i.v. injection was per-formed. No obvious toxicity or side effects were observed after the injections,except for NP12. After NP12 injection, mice had slow movement, and one outof five died within 24 h after the injection. For short-term toxicity study of Rx-HDL and Rx-PLGA-HDL, four injections in 8 d were given using a GW3965 doseof 10 mg/kg per injection. For the long-term treatment study, the mice re-ceived two i.v. injections of Rx-HDL (n = 12) per week for 6 wk at the dose of10 mg/kg GW3965 per injection or equal volume of PBS (n = 12). A separatecohort of female Apoe−/− mice with the same length of high-fat diet received6 wk of oral GW3965 (n = 6, 10 mg/kg GW3965, two times per week) or PBS(n = 5). No abnormal activities or deaths occurred during the treatmentregimen.

Micro-PET/CT Imaging. Apoe−/− atherosclerotic mice were injected with either89Zr-labeled Rx-HDL or Rx-PLGA-HDL at about 200 μCi per mouse (n = 5). At30 min and 24 h after the injection, the mice were imaged on an Inveon PET/CTscanner (Siemens Healthcare Global) under isoflurane-induced (Baxter Health-care) anesthesia. Whole-body static PET scans recorded a minimum of 50 millioncoincident events in ∼15 min. The energy and coincidence timing windows were350 keV to 700 keV and 6 ns, respectively. The image data were normalized to

correct for nonuniform PET response, dead-time count losses, positron branchingratio, and physical decay to the time of injection, but no attenuation, scatter, orpartial volume averaging correction was applied. The counting rates in thereconstructed images were converted to activity concentrations (percentage in-jected dose per gram of tissue) via a system calibration factor derived fromimaging a mouse-sized water-equivalent phantom containing 89Zr. Images wereanalyzed using ASIPro VMTM software (Concorde Microsystems). Activity con-centration was quantified by averaging the maximal values of at least 10 regionsof interest (ROIs) drawn on consecutive slices of the chosen organs.

Liver mRNA Expression Measurement. Total RNA was obtained from snap-frozen liver tissue and homogenized in TRIzol reagent (Ambion). The ho-mogenate was spun down to pellet tissue, and the aqueous TRIzol supernatantwas collected and processed using the Direct-Zol. RNA miniprep kit (ZymoResearch Corporation) was used for RNA purification. The RNA was thenreverse-transcribed using the Verso cDNA kit (Thermo Scientific) and dilutedusing RNase/DNase-free water. Quantitative real-time PCR was performedwith Taqman Gene Expression Master Mix (Applied Biosystems), and withTaqman primer/probe mixes for Abca1, Abcg1, or Srebp1c. Gene expressionwas normalized to 18S ribosomal RNA (rRNA) expression.

Aortic Macrophage mRNA Expression Measurement. The quality of RNA extractsfrom atherosclerotic plaques was measured by Agilent 2100 bioanalyzer(Agilent Technologies). High-quality [RNA integrity number (RIN) > 7] RNAsamples were amplified with a WT-Ovation Pico RNA amplification system(NuGen). The cDNA from the amplification reactions were used to run amicrofluidics-based mRNA profiling chip (BioMark; Fluidigm), which had highreproducibility. Hprt1 was used as the housekeeping gene. The following 24genes were measured: Abca1, Abcg1, Srebp1c, Nr1h2, Hr1h3, Ccl2, Ccr2, Icam1,Ifng, Il12a, Il1a, Il1b, Il6, Nos2,Mmp2,Mmp9, Tnfa, Vcam1, Ccr7, Cd206, Foxp3,Il10, Mrc2, and Tgif1. Twelve biological repeats were included for each gene.

Cellular Lipid Measurement with Flow Cytometry. Single cells were preparedfrom aortas from either Apoe−/− mice or wild-type C57BL/6 mice and thenstained with antibody mixtures as described in SI Materials and Methods, FlowCytometry. On the basis of a previous procedure (33), the stained cells werewashedwith PBS two times at room temperature and then incubatedwith 250 μLPBS containing 0.5 μg/mL BODIPY (D-3922; Molecular Probes) for 15 min. The cellswere washed with flow cytometry buffer two times and analyzed in a BectonDickinson LSRII flow cytometer.

Statistics. Data are presented as mean ± SE of mean (SEM) unless otherwisenoted. Two-tailed Student’s t test was used to calculate statistical significance.GraphPad Prism 5.0 for PC (GraphPad Software Inc.) was used for statisticalanalysis. P < 0.05 was regarded as significant; * denotes P value < 0.05, and** denotes P value < 0.01 in all figures in the paper.

More description of materials and methods is included in SupportingInformation.

ACKNOWLEDGMENTS.We thank the flow cytometry core facility, the preclinicalimaging center of Translational and Molecular Imaging Institute, and the quan-titative qPCR facility of Icahn School of Medicine at Mount Sinai. We thank thesmall animal imaging core facility, the genomic core facility, the molecular cytol-ogy core facility, and the radiochemistry and molecular imaging probe core ofMemorial Sloan Kettering Cancer Center. We thank Dr. Kaley Joyes for editingthe manuscript. This work was supported by National Institute of Health GrantsR01 HL118440, R01HL125703, and R01 CA155432 (to W.J.M.M.), R01 EB009638(to Z.A.F.), K25 EB016673 (to T.R.), R01HL123433 (to K.J.M.), and P30 CA008748-50(to Dr. Jason S. Lewis), as well as a Harold S. Geneen Charitable Trust Award(to Z.A.F.), European Framework Program 7 grant (FP7-Health 309820: Nano-Athero) and the Dutch network for Nanotechnology NanoNext NL (A.A. andG.S.), and a Netherlands Organisation for Scientific Research Vidi (to W.J.M.M.).

1. Fridman WH, Pagès F, Sautès-Fridman C, Galon J (2012) The immune contexture in

human tumours: Impact on clinical outcome. Nat Rev Cancer 12(4):298–306.2. Lehuen A, Diana J, Zaccone P, Cooke A (2010) Immune cell crosstalk in type 1 diabetes.

Nat Rev Immunol 10(7):501–513.3. Moore KJ, Sheedy FJ, Fisher EA (2013) Macrophages in atherosclerosis: A dynamic

balance. Nat Rev Immunol 13(10):709–721.4. Swirski FK, Nahrendorf M (2013) Leukocyte behavior in atherosclerosis, myocardial

infarction, and heart failure. Science 339(6116):161–166.5. Libby P (2002) Inflammation in atherosclerosis. Nature 420(6917):868–874.6. Heidt T, et al. (2014) Chronic variable stress activates hematopoietic stem cells. Nat

Med 20(7):754–758.7. Robbins CS, et al. (2013) Local proliferation dominates lesional macrophage accu-

mulation in atherosclerosis. Nat Med 19(9):1166–1172.

8. Galkina E, Ley K (2009) Immune and inflammatory mechanisms of atherosclerosis.Annu Rev Immunol 27:165–197.

9. Dutta P, et al. (2012) Myocardial infarction accelerates atherosclerosis. Nature487(7407):325–329.

10. Groot PH, et al. (2005) Synthetic LXR agonists increase LDL in CETP species. J Lipid Res46(10):2182–2191.

11. Kotokorpi P, et al. (2007) Physiological differences between human and rat pri-mary hepatocytes in response to liver X receptor activation by 3-[3-[N-(2-chloro-3-trifluoromethylbenzyl)-(2,2-diphenylethyl)amino]propyloxy]phenylacetic acid hydrochlo-ride (GW3965). Mol Pharmacol 72(4):947–955.

12. Tang J, et al. (2015) Inhibiting macrophage proliferation suppresses atheroscleroticplaque inflammation. Sci Adv 1(3):e1400223.

13. Duivenvoorden R, et al. (2014) A statin-loaded reconstituted high-density lipoproteinnanoparticle inhibits atherosclerotic plaque inflammation. Nat Commun 5:3065.

Tang et al. PNAS | Published online October 17, 2016 | E6739

ENGINEE

RING

IMMUNOLO

GYAND

INFLAMMATION

PNASPL

US

Dow

nloa

ded

by g

uest

on

Dec

embe

r 31

, 202

0

14. Sanchez-Gaytan BL, et al. (2015) HDL-mimetic PLGA nanoparticle to target athero-sclerosis plaque macrophages. Bioconjug Chem 26(3):443–451.

15. Cormode DP, et al. (2009) HDL as a contrast agent for medical imaging. Clin Lipidol4(4):493–500.

16. Nagano Y, Arai H, Kita T (1991) High density lipoprotein loses its effect to stimulateefflux of cholesterol from foam cells after oxidative modification. Proc Natl Acad SciUSA 88(15):6457–6461.

17. Matsumura Y (2008) Poly (amino acid) micelle nanocarriers in preclinical and clinicalstudies. Adv Drug Deliv Rev 60(8):899–914.

18. Torchilin VP (2005) Recent advances with liposomes as pharmaceutical carriers. NatRev Drug Discov 4(2):145–160.

19. Tall AR, Yvan-Charvet L, Terasaka N, Pagler T, Wang N (2008) HDL, ABC transporters,and cholesterol efflux: Implications for the treatment of atherosclerosis. Cell Metab7(5):365–375.

20. Tall AR, Yvan-Charvet L (2015) Cholesterol, inflammation and innate immunity. NatRev Immunol 15(2):104–116.

21. Schiener M, et al. (2014) Nanomedicine-based strategies for treatment of athero-sclerosis. Trends Mol Med 20(5):271–281.

22. Weissleder R, Ntziachristos V (2003) Shedding light onto live molecular targets. NatMed 9(1):123–128.

23. Galkina E, et al. (2007) CXCR6 promotes atherosclerosis by supporting T-cell homing,interferon-gamma production, and macrophage accumulation in the aortic wall.Circulation 116(16):1801–1811.

24. Galkina E, et al. (2006) Lymphocyte recruitment into the aortic wall before and duringdevelopment of atherosclerosis is partially L-selectin dependent. J Exp Med 203(5):1273–1282.

25. Weissleder R, Nahrendorf M, Pittet MJ (2014) Imaging macrophages with nano-particles. Nat Mater 13(2):125–138.

26. Hong C, Tontonoz P (2014) Liver X receptors in lipid metabolism: Opportunities fordrug discovery. Nat Rev Drug Discov 13(6):433–444.

27. Joseph SB, et al. (2002) Synthetic LXR ligand inhibits the development of athero-sclerosis in mice. Proc Natl Acad Sci USA 99(11):7604–7609.

28. Naik SU, et al. (2006) Pharmacological activation of liver X receptors promotes reversecholesterol transport in vivo. Circulation 113(1):90–97.

29. Deri MA, Zeglis BM, Francesconi LC, Lewis JS (2013) PET imaging with ⁸⁹Zr: From ra-diochemistry to the clinic. Nucl Med Biol 40(1):3–14.

30. Zhang SH, Reddick RL, Piedrahita JA, Maeda N (1992) Spontaneous hypercholesterol-emia and arterial lesions in mice lacking apolipoprotein E. Science 258(5081):468–471.

31. Wu KT, et al. (2016) Nonalcoholic fatty liver disease severity is associated with theratios of total cholesterol and triglycerides to high-density lipoprotein cholesterol.J Clin Lipidol 10(2):420–425.

32. Brunham LR, et al. (2006) Tissue-specific induction of intestinal ABCA1 expressionwith a liver X receptor agonist raises plasma HDL cholesterol levels. Circ Res 99(7):672–674.

33. Herber DL, et al. (2010) Lipid accumulation and dendritic cell dysfunction in cancer.Nat Med 16(8):880–886.

34. Zhao Y, et al. (2016) Augmenting drug-carrier compatibility improves tumour nano-therapy efficacy. Nat Commun 7:11221.

35. Mullard A (2015) PCSK9 inhibitors are go. Nat Rev Drug Discov 14(9):593.36. Sabatine MS, et al.; Open-Label Study of Long-Term Evaluation against LDL Choles-

terol (OSLER) Investigators (2015) Efficacy and safety of evolocumab in reducing lipidsand cardiovascular events. N Engl J Med 372(16):1500–1509.

37. Robinson JG, et al.; ODYSSEY LONG TERM Investigators (2015) Efficacy and safety ofalirocumab in reducing lipids and cardiovascular events. N Engl J Med 372(16):1489–1499.

38. Chames P, Van Regenmortel M, Weiss E, Baty D (2009) Therapeutic antibodies: Suc-cesses, limitations and hopes for the future. Br J Pharmacol 157(2):220–233.

39. Cook D, et al. (2014) Lessons learned from the fate of AstraZeneca’s drug pipeline: Afive-dimensional framework. Nat Rev Drug Discov 13(6):419–431.

40. Mieszawska AJ, et al. (2013) Synthesis of polymer-lipid nanoparticles for image-guided delivery of dual modality therapy. Bioconjug Chem 24(9):1429–1434.

41. Fredman G, et al. (2015) Targeted nanoparticles containing the proresolving peptideAc2-26 protect against advanced atherosclerosis in hypercholesterolemic mice. SciTransl Med 7(275):275ra20.

42. Choleris E, et al. (2007) Microparticle-based delivery of oxytocin receptor antisenseDNA in the medial amygdala blocks social recognition in female mice. Proc Natl AcadSci USA 104(11):4670–4675.

43. Palucka K, Banchereau J (2012) Cancer immunotherapy via dendritic cells. Nat RevCancer 12(4):265–277.

44. Hammad H, Lambrecht BN (2008) Dendritic cells and epithelial cells: Linking innateand adaptive immunity in asthma. Nat Rev Immunol 8(3):193–204.

45. Mukherjee G, Dilorenzo TP (2010) The immunotherapeutic potential of dendritic cellsin type 1 diabetes. Clin Exp Immunol 161(2):197–207.

46. Pérez-Medina C, et al. (2015) PET imaging of tumor-associated macrophages with89Zr-labeled high-density lipoprotein nanoparticles. J Nucl Med 56(8):1272–1277.

E6740 | www.pnas.org/cgi/doi/10.1073/pnas.1609629113 Tang et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 31

, 202

0